Spaces:
Running
on
Zero
Running
on
Zero
regex parsing
Browse files
app.py
CHANGED
@@ -7,6 +7,7 @@ subprocess.run(
|
|
7 |
)
|
8 |
import spaces
|
9 |
import gradio as gr
|
|
|
10 |
|
11 |
from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
|
12 |
from qwen_vl_utils import process_vision_info
|
@@ -18,15 +19,15 @@ from typing import Tuple
|
|
18 |
|
19 |
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
|
20 |
|
21 |
-
|
22 |
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
23 |
"Qwen/Qwen2.5-VL-7B-Instruct",
|
24 |
torch_dtype=torch.bfloat16,
|
25 |
attn_implementation="flash_attention_2",
|
26 |
device_map="auto",
|
27 |
)
|
28 |
-
processor = AutoProcessor.from_pretrained(
|
29 |
-
|
|
|
30 |
|
31 |
class GeneralRetrievalQuery(BaseModel):
|
32 |
broad_topical_query: str
|
@@ -36,6 +37,17 @@ class GeneralRetrievalQuery(BaseModel):
|
|
36 |
visual_element_query: str
|
37 |
visual_element_explanation: str
|
38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
39 |
|
40 |
def get_retrieval_prompt(prompt_name: str) -> Tuple[str, GeneralRetrievalQuery]:
|
41 |
if prompt_name != "general":
|
@@ -76,11 +88,9 @@ Generate the queries based on this image and provide the response in the specifi
|
|
76 |
|
77 |
return prompt, GeneralRetrievalQuery
|
78 |
|
79 |
-
|
80 |
# defined like this so we can later add more prompting options
|
81 |
prompt, pydantic_model = get_retrieval_prompt("general")
|
82 |
|
83 |
-
|
84 |
def _prep_data_for_input(image):
|
85 |
messages = [
|
86 |
{
|
@@ -109,7 +119,6 @@ def _prep_data_for_input(image):
|
|
109 |
return_tensors="pt",
|
110 |
)
|
111 |
|
112 |
-
|
113 |
@spaces.GPU
|
114 |
def generate_response(image):
|
115 |
inputs = _prep_data_for_input(image)
|
@@ -125,13 +134,20 @@ def generate_response(image):
|
|
125 |
generated_ids_trimmed,
|
126 |
skip_special_tokens=True,
|
127 |
clean_up_tokenization_spaces=False,
|
128 |
-
)
|
|
|
129 |
try:
|
130 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
131 |
except Exception:
|
132 |
gr.Warning("Failed to parse JSON from output")
|
133 |
-
return output_text
|
134 |
-
|
135 |
|
136 |
title = "ColPali Query Generator using Qwen2.5-VL"
|
137 |
description = """[ColPali](https://huggingface.co/papers/2407.01449) is a very exciting new approach to multimodal document retrieval which aims to replace existing document retrievers which often rely on an OCR step with an end-to-end multimodal approach.
|
|
|
7 |
)
|
8 |
import spaces
|
9 |
import gradio as gr
|
10 |
+
import re
|
11 |
|
12 |
from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
|
13 |
from qwen_vl_utils import process_vision_info
|
|
|
19 |
|
20 |
os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
|
21 |
|
|
|
22 |
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
|
23 |
"Qwen/Qwen2.5-VL-7B-Instruct",
|
24 |
torch_dtype=torch.bfloat16,
|
25 |
attn_implementation="flash_attention_2",
|
26 |
device_map="auto",
|
27 |
)
|
28 |
+
processor = AutoProcessor.from_pretrained(
|
29 |
+
"Qwen/Qwen2.5-VL-7B-Instruct",
|
30 |
+
)
|
31 |
|
32 |
class GeneralRetrievalQuery(BaseModel):
|
33 |
broad_topical_query: str
|
|
|
37 |
visual_element_query: str
|
38 |
visual_element_explanation: str
|
39 |
|
40 |
+
def extract_json_with_regex(text):
|
41 |
+
# Pattern to match content between code backticks
|
42 |
+
pattern = r'```(?:json)?\s*(.+?)\s*```'
|
43 |
+
|
44 |
+
# Find all matches (should typically be one)
|
45 |
+
matches = re.findall(pattern, text, re.DOTALL)
|
46 |
+
|
47 |
+
if matches:
|
48 |
+
# Return the first match
|
49 |
+
return matches[0]
|
50 |
+
return None
|
51 |
|
52 |
def get_retrieval_prompt(prompt_name: str) -> Tuple[str, GeneralRetrievalQuery]:
|
53 |
if prompt_name != "general":
|
|
|
88 |
|
89 |
return prompt, GeneralRetrievalQuery
|
90 |
|
|
|
91 |
# defined like this so we can later add more prompting options
|
92 |
prompt, pydantic_model = get_retrieval_prompt("general")
|
93 |
|
|
|
94 |
def _prep_data_for_input(image):
|
95 |
messages = [
|
96 |
{
|
|
|
119 |
return_tensors="pt",
|
120 |
)
|
121 |
|
|
|
122 |
@spaces.GPU
|
123 |
def generate_response(image):
|
124 |
inputs = _prep_data_for_input(image)
|
|
|
134 |
generated_ids_trimmed,
|
135 |
skip_special_tokens=True,
|
136 |
clean_up_tokenization_spaces=False,
|
137 |
+
)[0]
|
138 |
+
|
139 |
try:
|
140 |
+
# Try to extract JSON from code block first
|
141 |
+
json_str = extract_json_with_regex(output_text)
|
142 |
+
if json_str:
|
143 |
+
parsed = json.loads(json_str)
|
144 |
+
return json.dumps(parsed, indent=2)
|
145 |
+
# If no code block found, try direct JSON parsing
|
146 |
+
parsed = json.loads(output_text)
|
147 |
+
return json.dumps(parsed, indent=2)
|
148 |
except Exception:
|
149 |
gr.Warning("Failed to parse JSON from output")
|
150 |
+
return output_text
|
|
|
151 |
|
152 |
title = "ColPali Query Generator using Qwen2.5-VL"
|
153 |
description = """[ColPali](https://huggingface.co/papers/2407.01449) is a very exciting new approach to multimodal document retrieval which aims to replace existing document retrievers which often rely on an OCR step with an end-to-end multimodal approach.
|